1. Technical Field
The present invention relates to an article pickup apparatus for picking up randomly piled articles by a robot.
2. Description of the Related Art
An article pickup apparatus for picking up randomly piled articles by a robot is known. The article pickup apparatus visually recognizes a loaded state of articles based on a two-dimensional image of the articles captured by a camera or a three-dimensional point set acquired using a three-dimensional measurement device (See Japanese Laid-open Patent Publications No. 2011-179909 and No. 2011-093058).
Japanese Laid-open Patent Publication No. 2011-179909 discloses a position and posture measurement apparatus which estimates a position and a posture of a target object using a distance image of the target object. The position and posture measurement apparatus according to the known technique compares information indicating a distance between an object model and an object area with information indicating a distance between a partial area, which is not set as the object area in a partial area group, and the object model, and updates the object area by adding a new partial area to the object area according to the comparison result, in order to estimate the position and the posture of the target object by associating a group of three-dimensional points of the updated object area with the object model.
Japanese Laid-open Patent Publication No. 2011-093058 discloses a technique for extracting a grip area including an object to be gripped by a grip mechanism of a robot from three-dimensional information of a supply unit on which an object whose position and posture are unknown is provided. According to the known technique, the three-dimensional information of the supply unit is generated from three-dimensional measurement data of the supply unit for an object and extracts, using a grip area stored in advance, which includes a grip portion area and a grip mechanism area, an area where the object exists in an entire grip portion area and does not exists in the grip mechanism area, as a grippable area from the three-dimensional information.
However, if a position and a posture of an article or a grip position and posture of a hand is detected based on visual information acquired using a visual sensor as in the cases of the known techniques described in Japanese Laid-open Patent Publications No. 2011-179909 and No. 2011-093058, there is always a risk of false detection, detection failure, or a detection error. For example, in the case where an article is gripped by a chuck hand including a pair of claws, when a center of the hand does not match with a center of the article, only one claw of the hand comes into contact with the article, which makes it difficult to close the hand. In this case, a gap is formed between the other claw and the article, and thus it is likely that the hand fails to hold the article. If the hand is forcibly closed, a robot or the hand is subjected to an overload, or the article may be damaged by an excess force.
Japanese Laid-open Patent Publication No. 2000-263481 discloses a technique in which a sensor for detecting an abnormal operation is installed in a manipulator, and a pickup operation is stopped to temporarily retract the manipulator when the sensor detects an abnormality of the operation during a pickup operation of an article by the manipulator, in order to detect the article again by a visual recognition device.
However, in the known technique described in Japanese Laid-open Patent Publication No. 2000-263481, the manipulator needs to retract every time an abnormality is detected. Therefore, the pickup operation may be terminated even if there is an article which can be actually picked up. For example, in order to avoid interference of the hand with an article in the surroundings or a container, a center of the open hand may be intentionally shifted from a center of an article. However, according to the technique disclosed in Japanese Laid-open Patent Publication No. 2000-263481, when the center of the hand is shifted from the center of the article, the abnormality is detected, and as a result, the pickup operation of the article is stopped. This means that even if there is an article which can be potentially picked up, the pickup process is stopped.
Therefore, an article pickup apparatus is needed which can stably pick up an article based on visual information acquired by a visual sensor.